feng-hirst-service

This docker container allows you to build, install and run the Feng/Hirst discourse parser (Feng and Hirst 2014) in a docker container with an added REST API. Instead of the original code, this service uses my fork which applies some patches, is easier to dockerize and produces an output format that is easier to parse (e.g. by discoursegraphs or the rst-converter-service).

build

docker build -t feng-hirst-service .

run

docker run -p 8000:8000 -ti feng-hirst-service

Usage Examples

CURL

$ cat input_short.txt
Although they didn't like it, they accepted the offer.

$ curl -X POST -F "input=@input_short.txt" http://localhost:8000/parse
ParseTree('Contrast[S][N]', ["Although they did n't like it ,", 'they accepted the offer .'])

Javascript

This works in Chrome, but not in Firefox:

>>> var xhr = new XMLHttpRequest();

>>> xhr.open("POST", "http://localhost:8000/parse")

>>> var data = new FormData();
>>> data.append('input', 'Altough they didn\'t like him, they accepted the offer.');

>>> xhr.send(data);
>>> console.log(xhr.response);
ParseTree('Background[S][N]', ["Altough they did n't like him ,", 'they accepted the offer .'])

Citation

If you use the Feng/Hirst RST parser in your academic work, please cite the following paper:

Vanessa Wei Feng and Graeme Hirst, 2014.
A Linear-Time Bottom-Up Discourse Parser with Constraints and Post-Editing.
In Proceedings of the 52th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-2014), Baltimore, USA.

For more technical details, see:

Vanessa Wei Feng and Graeme Hirst, 2014.
Two-pass Discourse Segmentation with Pairing and Global Features.
arXiv:1407.8215v1.